《计算机应用研究》|Application Research of Computers

关联性动态加权的协同过滤推荐

Relevance dynamic weighted collaborative filtering recommendation

免费全文下载 (已被下载 次)  
获取PDF全文
作者 王剑,余青松
机构 华东师范大学 计算机科学与软件工程学院,上海 200333
统计 摘要被查看 次,已被下载
文章编号 1001-3695(2019)11-005-3230-03
DOI 10.19734/j.issn.1001-3695.2018.05.0306
摘要 利用传统的协同过滤(CF)算法进行推荐时,由于用户评分矩阵比较稀疏,直接得到的用户或者项目之间的相似度相对而言可信度就比较低。为了解决这个问题,在传统的协同过滤基础上,引入项目与项目之间的关联性,通过在项目的类别标签和二部图的方法之间构建动态权重因子来融合这两种关联,形成非对等关联性关系,并做出用户对项目的评分预测,从而解决评分矩阵过于稀疏的问题。研究结果表明,相比于传统方法中使用对等相似度关系以及固定权值的方法,通过动态权重融合关联性形成非对等的关系的方法,更贴合生活实际,并且有更好的推荐效果。
关键词 协同过滤; 评分矩阵; 稀疏; 动态权重因子; 非对等关联性
基金项目
本文URL http://www.arocmag.com/article/01-2019-11-005.html
英文标题 Relevance dynamic weighted collaborative filtering recommendation
作者英文名 Wang Jian, Yu Qingsong
机构英文名 College of Computer Science & Software Engineering,East China Normal University,Shanghai 200333,China
英文摘要 When the traditional collaborative filtering algorithm is used for recommendation, the credibility based on similarity between users or items directly obtained is relatively low due to the sparsity of the user rating matrix. In order to solve this problem, this paper introduced the relevance between projects on the basis of traditional collaborative filtering. It established the association in a non-reciprocal condition by building a dynamic weighting factor between the project's category label and bipartite graph approach, and the user's rating of the project would be predicted. As a result, compared with the method of using the equivalent similarity and the fixed weight value in the traditional method, the method of using non-equivalent relationship formed by the dynamic weights is more in line with the reality of life and has a better recommendation effect.
英文关键词 collaborative filtering; rating matrix; sparse; dynamic weighting factor; non-equivalent relationship
参考文献 查看稿件参考文献
 
收稿日期 2018/5/25
修回日期 2018/7/9
页码 3230-3232,3249
中图分类号 TP301.6
文献标志码 A